Consensus Clustering - Related Work

Related Work

1. Clustering ensemble (Strehl and Ghosh): They considered various formulations for the problem, most of which reduce the problem to a hyper-graph partitioning problem. In one of their formulations they considered the same graph as in the correlation clustering problem. The solution they proposed is to compute the best k-partition of the graph, which does not take into account the penalty for merging two nodes that are far apart.

2. Clustering aggregation (Fern and Brodley): They applied the clustering aggregation idea to a collection of soft clusterings they obtained by random projections. They used an agglomerative algorithm and did not penalize for merging dissimilar nodes.

3. Fred and Jain: They proposed to use a single linkage algorithm to combine multiple runs of the k-means algorithm.

4. Dana Cristofor and Dan Simovici: They observed the connection between clustering aggregation and clustering of categorical data. They proposed information theoretic distance measures, and they propose genetic algorithms for finding the best aggregation solution.

5. Topchy et al.: They defined clustering aggregation as a maximum likelihood estimation problem, and they proposed an EM algorithm for finding the consensus clustering.

6. Abu-Jamous et al.: They proposed their binarization of consensus partition matrices (Bi-CoPaM) method to enhance ensemble clustering in two major aspects. The first is to consider clustering the same set of objects by various clustering methods as well as by considering their features measured in multiple datasets; this seems perfectly relevant in the context of microarray gene expression clustering, which is the context they initially proposed the method in. The second aspect is the format of the final result; based on the consistency of inclusion of a data object in the same cluster by the multiple single clustering results, they allowed any single data object to have any of the three eventualities; to be exclusively assigned to one and only one cluster, to be unassigned from all clusters, or to be simultaneously assigned to multiple clusters at the same time. They made it possible to produce, in a perfectly tunable way, wide overlapping clusters, tight specific clusters, as well as complementary clusters. Therefore, they proposed their work as a new paradigm of clustering rather than merely a new ensemble clustering method.